SPOP implemented. Hash table resizing for Sets and Expires too. Changed the resize...

[redis.git] / TODO

diff --git a/TODO b/TODO
index 89b67c5700fb9380f32c25d4f5343b4fabeea021..e37e50f22bf34fddcbee4405cfae4e3e9fa60eac 100644 (file)
--- a/TODO
+++ b/TODO
@@ -1,30 +1,33 @@
 BEFORE REDIS 1.0.0-rc1
 
 BEFORE REDIS 1.0.0-rc1
 

-- What happens if the saving child gets killed instead to end normally? Handle this.
-- Make sinterstore / unionstore / sdiffstore returning the cardinality of the resulting set.
-- Add a new field as INFO output: bgsaveinprogress
-- Remove max number of args limit
-- GETSET
-- network layer stresser in test in demo, make sure to set/get random streams of data and check that what we read back is byte-by-byte the same.
-- maxclients directive
-- check 'server.dirty' everywere
-- config parameter to change the name of the DB file
-- replication automated tests
-- an external tool able to perform the 'difference' between two Redis servers. It's like 'diff', but against Redis servers, and the output is the set of commands needed to turn the first server into the second, suitable to be sent via netcat.
-   $ ./redis-diff 192.168.1.1 192.168.1.2 > diff.txt
-   $ cat diff.txt | nc 192.168.1.1 6379
-   $ ./redis-diff 192.168.1.1 192.168.1.2
-   $ # No output now the servers are identical
+ * Add number of keys for every DB in INFO
+ * Resize the expires and Sets hash tables if needed as well? For Sets the right moment to check for this is probably in SREM
+ * check 'server.dirty' everywere. Make it proprotional to the number of objects modified.
+ * Cover most of the source code with test-redis.tcl
+ * Remove tmp-.... files when saving child exits in the wrong way, to do so use tmp-pid.rdb as filename so that the parent can rebuild the file name just from the child pid.

 
 

-This command should be smart and don't use too much memory, that is, take two connections at the same time against the two servers and perform the comparison key by key. Probably the initial "KEYS *" is unavoidable.
+AFTER 1.0 stable release

 
 

-- Shutdown must kill other background savings before to start saving. Otherwise the DB can get replaced by the child that rename(2) after the parent for some reason.
-- Add missing commands in documentation
-- Document replication
-- Objects sharing configuration, add the directive "objectsharingpool <size>"
-- Make sure to convert all the fstat() calls to 64bit versions.
-- SINTERCOUNT, SUNIONCOUNT, SDIFFCOUNT
+ * Consistent hashing implemented in all the client libraries having an user base
+ * SORT: Don't copy the list into a vector when BY argument is constant.
+ * SORT ... STORE keyname. Instead to return the SORTed data set it into key.
+ * Profiling and optimization in order to limit the CPU usage at minimum
+ * Write the hash table size of every db in the dump, so that Redis can resize the hash table just one time when loading a big DB.
+ * Elapsed time in logs for SAVE when saving is going to take more than 2 seconds
+ * LOCK / TRYLOCK / UNLOCK as described many times in the google group
+ * Replication automated tests
+ * some kind of sorted type, example:
+        ZADD mykey foo 100
+        ZADD mykey bar 50
+   ZRANGE mykey 0 1 => bar foo
+   This is able to take elements sorted because a binary tree is used to store
+   the elements by 'score', with the actual value being the key. On the other
+   side the type also takes an hash table with key->score mapping, so that when
+   there is an update we lookup the current score and can traverse the tree.
+ * BITMAP type
+ * LRANGE 4 0 should return the same elements as LRANGE 0 4 but in reverse order (only if we get enough motivated requests about it)
+ * zmalloc() should avoid to add a private header for archs where there is some other kind of libc-specific way to get the size of a malloced block.

 
 FUTURE HINTS
 
 
 FUTURE HINTS
 

-- if in-memory values compression will be implemented, make sure to implement this so that addReply() is able to handle compressed objects, just creating an uncompressed version on the fly and adding this to the output queue instead of the original one. When insetad we need to look at the object string value (SORT BY for example), call a function that will turn the object into an uncompresed one.
+- In memory compression: if in-memory values compression will be implemented, make sure to implement this so that addReply() is able to handle compressed objects, just creating an uncompressed version on the fly and adding this to the output queue instead of the original one. When insetad we need to look at the object string value (SORT BY for example), call a function that will turn the object into an uncompresed one.